Buying LED Drivers from China: What the Datasheet Won't Tell You

The first time I bought LED fixtures for a project in the north of England, I lost about six percent of them in the first eighteen months. All of it came back to the driver, not the LEDs. The chips were fine. The driver was a constant-current supply built to hit a price, sized for a lab bench, and installed in a sealed can where it sat at 70 degrees C all day for years. That lesson cost me less than a proper lab test would have, but only barely.

Constant current is the default, and it sets everything else

Most fixtures you will buy out of Guangdong use a constant-current driver, usually 300mA, 350mA, 500mA or 700mA on the low end and 1050mA to 1500mA for higher-power COBs. The current is what determines the light output and the heat load. If your supplier changes the driver from 350mA to 500mA because they had stock that week, the COB gets hotter, the lumen output shifts, and the lifetime curve you were quoted is now fiction. I now write the drive current, the voltage window and the acceptable driver model into the purchase spec and treat any substitution as a change request.

Constant-voltage supplies (12V or 24V) are the other family, and they are what you want for strips and simple linear fittings. Mixing them with a current-driven COB produces either no light or a very short life. Ask which family the driver is in before you discuss anything else.

Efficiency is only real at operating temperature

A driver advertised at 91 percent efficiency was measured at 25 degrees ambient with a resistive load. In the enclosure, at 60 degrees case temperature, the same sample measured about 86 percent. That gap is not a lie exactly, it is a test condition, and the heat it creates is what kills the electrolytic capacitors. Put the operating condition in writing: efficiency at nominal load and at the case temperature the fixture will actually see.

While you are there, ask for the derating curve. A good driver holds its rated output to 50 or 60 degrees case and then steps down. A cheap one just keeps going until it stops, and the stopping is what you will see in the field as dead fixtures in the third summer.

The capacitors are the specification, whether they admit it or not

Electrolytic capacitors are the single biggest predictor of driver life. Known brands like Rubycon, Nichicon, Nippon Chemi-Con and Aishi cost more, and suppliers substitute generic parts to reach a price point. The realistic response is not to demand a brand name for a 5 dollar driver, but to ask for the capacitor brand and the rated hours at 105 degrees, and to decide whether the answer is good enough for your application. For a fixture that runs eight hours a day this matters far less than for one running 24/7 in a factory.

Also look at whether the driver is potted and with what. Fully potted drivers survive condensation and vibration much better. Half-potted or unpotted ones are common and are not automatically bad, but if the fixture is going outdoors or into a washdown area, the potting compound and the gasket around the cable entry matter more than a brand name.

Ripple and flicker: the cheap drivers lose you the health-conscious buyer

Cheap drivers with undersized smoothing capacitors produce output ripple, which shows up as flicker, which shows up in phone camera videos and in complaints. Flicker percent at 100Hz is the number you want, measured at nominal load. Anything above roughly 10 percent starts to be noticeable in video; above 30 percent some people report headaches. A driver that passes EN 61000-3-3 does not automatically pass on flicker, because these are different tests.

For dimmable fixtures, also ask about the minimum dim level and the dimming curve. A driver that claims 1 to 100 percent dimming often holds steady down to 20 percent and then drops out, or worse, dims linearly in perceived light so the bottom third of the range looks useless.

Surge protection is the one number you should not skip

A 2kV surge rating is the minimum sensible for mains-connected lighting in most of Europe, and 4kV or higher for anything with long cable runs or overhead distribution upstream. Many budget drivers have an MOV and no gas discharge tube, which is adequate for 2kV and not for 6kV. If your project is a car park or an industrial site, the surge number and the test standard (EN 61000-4-5, and which port and coupling mode) should be in the datasheet, not discovered later.

Dimming protocol: your controls guy needs to know this early

Phase cut (leading or trailing edge), 0-10V, DALI-2, DMX and Zigbee are all in use. The problem I keep running into is the fixture supplier who has a DALI logo on the driver but has never tested it against a real DALI controller on a multi-channel bus. Ask which certification body tested the driver and ask for the protocol statement, then have your own controller test a sample before the order goes in. DALI is unforgiving about addressing, and a driver that authenticates but drops off the bus at address 32 will make the commissioning day miserable.

Certificates belong to the driver, not the fixture

This is where a lot of purchase orders quietly fall apart. The LED fixture carries a CE mark because the assembler declared it. The driver inside carries its own safety certificate, its own EN 61347 file, its own CE declaration. Those are not guaranteed to match. A fixture maker can buy a certified driver from a reputable supplier and build a compliant fixture around it, or it can buy a driver that has a certificate number printed on the label but whose file does not exist.

Check the driver's model number on the certificate against the label on the sample. Check that the certificate covers the current range you are buying, since a certificate for the 350mA version may not cover the 700mA variant. Then keep the driver certificate in your technical file, because the person auditing the fixture will ask for it and a screenshot of the supplier's webpage will not do.

A first order that actually tells you something

Buy at least 50 units of the exact configuration, not a sample of three. Run a 72 hour burn-in in your own facility, or negotiate a burn-in in the factory where you or an inspector can be present. Measure case temperature at the start and at the end. If the number moves upward by more than a handful of degrees, the thermal design is marginal and the batch will get worse as the ambient rises. I also stop paying for fancy lab tests on the first order and instead spend that money on a burn-in and an in-situ temperature check, which has caught more real problems for me than any datasheet ever has.

The decent driver suppliers in Guangdong are quite good. They will send you thermal data, a derating chart and the actual capacitor list if you ask. The ones that answer with a photo of the driver and a price are telling you their entire value proposition at the same time.